[Group 2] Predict Churning Customers

Groups

• 朱宇方, 107703035
• 林子恩, 107703046
• 郭家瑜, 107308016
• 穆永綸, 109971003

Goal

Predict if a customer is going to stop using the credit card services.

Demo

You should provide an example commend to reproduce your result

Rscript  code/merged.R --input data/BankChurners.csv --output results/performance.csv

• any on-line visualization
  • Url：https://wutever0017.shinyapps.io/bankchurner_v1/

Folder organization and its related information

docs

• Your presentation, docs/1101DS_Group2.pptx

data

• Source : Credit Card customers from Kaggle
• Url : https://www.kaggle.com/sakshigoyal7/credit-card-customers
• Input format: CSV file
• Any preprocessing?
  • Handle missing data: Income_Category
  • SMOTE

code

• Which method do you use?

  • Models for Classification Task
  • SVM
  • NaiveBayes
  • XGBoost
  • Random Forest

• What is a null model for comparison?

  • Accuracy of Null Model : 0.8393

results

• Which metric do you use
  • precision, recall, R-square
• Is your improvement significant?

type	training	testing
NaiveBayes	0.81	0.81
SVM	0.94	0.92
XGBoost	0.99	0.95
Random Forest	0.95	0.95

References

• Code/implementation which you include/reference

  • SMOTE

  • Mode

  • Naive Bayes

  • SVM

  • One-hot Encoding

  • Xgboost

  • ggplot2

  • Random Forest

• Packages you use

  • library(e1071)

  • library(caTools)

  • library(caret)

  • library(DMwR)

  • library(randomForest)

  • library(tidyverse)

  • library(xgboost)

  • library(party)

  • library(shiny)

  • library(ggbiplot)

  • library(MASS)

  • library(dplyr)

  • library(ggplot2)

  • library(RColorBrewer)

  • library(gridExtra)

  • library(cowplot)

• Related publications

  • Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.